why torch-to-iree pass generate func named main$async automatically?

[bailuan]

What happened?

When I try to use iree-opt --torch-to-iree to transform ir from torch to linalg. I found my function named main becomes main$async automatically. What should I do if I just want a function named main but not a async function named main$async?

Steps to reproduce your issue

1. Go to '...'
2. Click on '....'
3. Scroll down to '....'
4. See error

What component(s) does this issue relate to?

MLIR

Version information

No response

Additional context

No response

[ScottTodd]

If you just want to convert from torch to linalg, use torch-mlir-opt from https://github.com/llvm/torch-mlir. IREE's use of torch-mlir is in service of IREE compilation and may introduce other dialects as needed for that purpose.

The --torch-to-iree pass should be creating both a function named main and a function named main$async. That's documented here: https://github.com/iree-org/iree/blob/main/compiler/plugins/input/Torch/InputConversion/FuncConversion.cpp and tested here: https://github.com/iree-org/iree/blob/main/compiler/plugins/input/Torch/InputConversion/test/func_conversion.mlir.

[bailuan]

THX, another question please, what should I do if I just want use python script to transform IR but not compile from source to use torch-mlir-opt?

[bailuan]

I have tried to conpile from source by using command line suggested: cmake -GNinja -Bbuild \ -DCMAKE_BUILD_TYPE=Release \ -DPython3_FIND_VIRTUALENV=ONLY \ -DLLVM_ENABLE_PROJECTS=mlir \ -DLLVM_EXTERNAL_PROJECTS="torch-mlir" \ -DLLVM_EXTERNAL_TORCH_MLIR_SOURCE_DIR="$PWD" \ -DMLIR_ENABLE_BINDINGS_PYTHON=ON \ -DLLVM_TARGETS_TO_BUILD=host \ externals/llvm-project/llvm, but not find the torch-mlir-opt in build/bin/.

[ScottTodd]

THX, another question please, what should I do if I just want use python script to transform IR but not compile from source to use torch-mlir-opt?

The IREE Python bindings expose some of the MLIR Python bindings (https://mlir.llvm.org/docs/Bindings/Python/) via from iree.compiler import ir. See https://iree.dev/reference/bindings/python/#installing-iree-packages for installation instructions. Upstream MLIR and torch-mlir also build some subset of the python bindings into packages.

I have tried to conpile from source by using command line suggested: cmake -GNinja -Bbuild \ -DCMAKE_BUILD_TYPE=Release \ -DPython3_FIND_VIRTUALENV=ONLY \ -DLLVM_ENABLE_PROJECTS=mlir \ -DLLVM_EXTERNAL_PROJECTS="torch-mlir" \ -DLLVM_EXTERNAL_TORCH_MLIR_SOURCE_DIR="$PWD" \ -DMLIR_ENABLE_BINDINGS_PYTHON=ON \ -DLLVM_TARGETS_TO_BUILD=host \ externals/llvm-project/llvm, but not find the torch-mlir-opt in build/bin/.

Please direct torch-mlir support questions to torch-mlir. Did you build after configuring? See https://github.com/llvm/torch-mlir/blob/main/docs/development.md#build-commands

[bailuan]

ok, I will try this.